Use tiktoken for faster token count

Author: intitni

Tool/Sources/OpenAIService/Memory/AutoManagedChatGPTMemoryStrategy/AutoManagedChatGPTMemoryGoogleAIStrategy.swift

@@ -7,13 +7,18 @@ extension AutoManagedChatGPTMemory {
         let configuration: ChatGPTConfiguration
 
         func countToken(_ message: ChatMessage) async -> Int {
-            guard let model = configuration.model else {
-                return 0
-            }
-            let aiModel = GenerativeModel(name: model.info.modelName, apiKey: configuration.apiKey)
-            if message.isEmpty { return 0 }
-            let modelMessage = ModelContent(message)
-            return (try? await aiModel.countTokens([modelMessage]).totalTokens) ?? 0
+            (await OpenAIStrategy().countToken(message)) + 10
+            // Using local tiktoken instead until I find a faster solution.
+            // The official solution requires sending a lot of requests when adjusting the prompt.
+            // adding 10 just incase.
+            
+//            guard let model = configuration.model else {
+//                return 0
+//            }
+//            let aiModel = GenerativeModel(name: model.info.modelName, apiKey: configuration.apiKey)
+//            if message.isEmpty { return 0 }
+//            let modelMessage = ModelContent(message)
+//            return (try? await aiModel.countTokens([modelMessage]).totalTokens) ?? 0
         }
 
         func countToken<F>(_: F) async -> Int where F: ChatGPTFunction {